INTRODUCTION TO POWER MOSFETS
What is a POWER MOSFET?
IR HEXFET, Motorola TMOS, Ixys HiPerFETs
and MegaMOS, Siemens SIPMOS power transistors, Advanced Power Technology
Whether the process is called VMOS, TMOS or DMOS they are all
We all know how to use a diode to implement a switch.
we can only switch with it, not gradually control the signal flow.
Furthermore, a diode acts as a switch depending on the direction of signal
flow; we can’t program it to pass or block a signal.
For such applications
involving either "flow control" or programmable on/off switching we need a
3-terminal device and Bardeen & Brattain heard us and "invented" (almost by
accident, like many other great discoveries!) the bipolar transistor.
Structurally it is implemented with only two junctions back-to-back (no big
deal; we were probably making common cathodes - same structure -long before
But functionally it is a totally different device which acts like a
"faucet" controlling the flow of emitter current -and the "hand" manipulating
the faucet is the base current.
A bipolar transistor is therefore a
The Field Effect Transistor (FET), although
structurally different, provides the same "faucet" function.
the FET is voltage-controlled; one doesn’t need base current but voltage to
exercise flow control.
The bipolar transistor was born in 1947; the FET (at
least the concept) came soon after, in 1948 from another pair of illustrious
parents: Shockley and Pearson.
The terminals are called DRAIN instead of
COLLECTOR, GATE instead of BASE and SOURCE instead of EMITTER to differentiate
it from his older bipolar "cousin".
The FET comes in two major variants,
optimised for different types of applications: the JFET (junction FET) used in
small-signal processing and the MOSFET (metal-oxide-semiconductor FET) mainly
used in linear or switching power applications.
Why Did They Need to Invent
the Power MOSFET?
When scaled-up for power applications the bipolar
transistor starts showing some annoying limitations.
Sure, you can still
find it in your washing machine, in your air conditioner and refrigerator but
these are "low" power applications for us, the average consumer, who can
tolerate a certain degree of inefficiency in his appliances.
still used in some UPSs, motor controls or welding robots but their usage is
practically limited to less than 10kHz and they are rapidly disappearing from
the "technology edge" applications where overall efficiency is the "key"
parameter (SMPS (switch-mode power supplies), sophisticated motor controls,
converters, to name a few).
Being a bipolar device, the transistor relies on
the minority carriers injected in the base to "defeat" recombination and be
re-injected in the collector.
In order to sustain a large collector current
we want to inject many of them in the base from the emitter side and, if
possible, recuperate all of them at the base/collector boundary (meaning that
recombination in the base should be kept at a minimum).
But this means that
when we want the transistor switched off, there will be a considerable amount of
minority carriers in the base with a low recombination factor to be taken care
of before the switch can close - in other words the stored charge problem
associated with all minority carrier devices limiting the maximum operating
The major advantage of the FET now comes to light: being a majority
carrier device there is no stored minority charge therefore it can work at much
The switching delays characteristic to mosfets are rather
a consequence of the charging and discharging of the parasitic
One may say: I see the need for a fast switching mosfet in high
frequency applications but why should I use such device in my relatively slow
The answer is straightforward: improved
The device sees both high current and high voltage during the
interval in which switching occurs; a faster device will therefore experience
proportionally less energy loss.
In many applications this advantage alone
more than compensates for the slightly higher conduction losses associated with
higher voltage mosfets: SMPS (switch-mode power supplies) operating beyond 150
kHz would not be possible without them.
The bipolar transistor is current
driven; in fact the more current we want to drive, the more current we need to
supply to the base because the gain (ratio of the collector and base currents)
drops significantly as the collector current (IC) increases. One consequence is
that the bipolar transistor starts dissipating significant control power,
reducing the overall efficiency of the circuitry.
To make things worse this
drawback is accentuated at higher operating temperatures. Another consequence is
the need for rather complicated base drive circuitry capable of fast current
sourcing and sinking.
Not the (MOS)FET; this device has practically zero
current consumption in the gate; even at 125°C the typical gate current stays
below 100 nA.
Once the parasitic capacitances are charged, only the very low
leakage currents have to be provided by the drivers.
Add to this the circuit
simplicity resulting from driving a device with voltage rather than current and
you’ll spot another reason why the (MOS)FET is so appealing to the design
Another major advantage is the non-existence of a secondary
breakdown mechanism. Try to block a lot of power with a bipolar transistor;
local defects unavoidable in any semiconductor structure will act to concentrate
the current, the result will be localised heating of the silicon.
temperature coefficient of the resistivity is negative the local defect will act
as low resistance current path, directing even more current into it, self
heating even more until non-reversible destruction occurs.
The MOSFET has a
positive resistivity thermal coefficient.
On one hand this can be perceived
as the disadvantage of an increased RDS(on) at elevated temperatures - this
important parameter roughly doubles between 25° C and 125° C due to carrier
On the other hand this same phenomenon brings a
significant advantage: any defect trying to act as described above would
actually divert current from it -one would have "cooling-spots" instead of the
"hot-spots" characteristic to bipolar devices!
An equally important
consequence of this self-cooling mechanism is the ease of paralleling MOSFETS to
boost-up the current capability of a device.
Bipolar transistors are very
sensitive to paralleling; precautions (emitter ballasting resistors, fast
response current-sensing feedback loops) have to be taken for equal sharing of
currents, otherwise the device with the lowest saturation voltage would divert
most of the current, overheating as described above and ultimately resulting in
Not the MOSFET; they can be paralleled with no other
precautions than design insured circuit symmetry and balancing the gates so they
open equally allowing the same amount of current in all transistors.
extra bonus is that even if the gates are not balanced and the channels have
different degrees of opening, this would still result in a steady state
condition with some drain currents being slightly larger than others. A useful
feature appealing to the design engineer is a consequence of the unique
structure of the mosfet (see Fig 3 for a more detailed description): the
"parasitic" body-diode formed between source and drain.
Although it is not
optimised for fast switching or low conduction loss, it acts as a clamping diode
in inductive load switching applications at no extra cost.
The basic idea of a JFET
(fig.1) is to control the current flowing from source to drain by modulating
(pinching) the cross-sectional area of the Drain-Source channel.
achieved using a reverse biased junction as gate; its (reverse) voltage
modulates the depleted region consequently pinching the channel and increasing
its resistance by reducing its section. With no voltage applied to the gate the
channel resistance is at its lowest value and maximum drain current flows
through the device.
As gate voltage is
increased the two depleted region fronts advance, reducing the drain current by
increasing the channel resistance until total pinch-off occurs when the two
The channel is now severed and no more current
MOSFET uses a different type of gate mechanism exploiting the properties of the
MOS capacitor. By varying the value and the polarity of the bias applied to the
top electrode of a MOS structure one can drive the silicon underneath it into
enhancement all the way to inversion.
the simplified structure of an N-channel MOSFET. It is called a vertical, double
diffused structure and starts with a heavily concentrated n substrate in order
to minimise the bulk portion of the channel resistance.
An n- epi layer is
grown on it and two successive diffusions are made, a p- zone in which proper
bias will generate the channel and an n+ into it defining the source. Next the
thin, high quality gate oxide is grown followed by the phosphorous-doped
poly-silicon thus forming the gate.
Contact windows are opened on top
defining the source and the gate terminals while the whole bottom of the wafer
makes the drain contact.
With no bias on the gate the n+ source and the n
drain are separated by the p zone and no current flows (transistor is
turned-off). If a positive bias is applied to the gate the minority carriers in
the p zone (electrons) are attracted to the surface underneath the gate
As the bias increases more electrons are being confined to this small
space, the local "minority" concentration becomes larger than the hole (p)
concentration and "inversion" occurs (meaning that the material immediately
under the gate turns from p- to n-type). Now an n "channel" is formed in the p
material right under the gate structure connecting the source to the drain;
current can now flow.
Like in the case of the JFET (although the physical
phenomenon is different) the gate (by means of its voltage bias) controls the
flow of current between the source and the drain.
There are many MOSFET
manufacturers and almost everyone has his own process optimisation and his
International Rectifier pioneered the HEXFET, Motorola builds
TMOS, Ixys fabricates HiPerFETs and MegaMOS, Siemens has the SIPMOS family of
power transistors and Advanced Power Technology the Power MOS IV, to name a few.
Whether the process is called VMOS, TMOS or DMOS it has a horizontal gate
structure and vertical current flow past the gate.
The power MOSFET is
nothing else but a structure containing a multitude of "cells" like the one
described in fig.2 connected in parallel.
And like any paralleling of
identical resistors the equivalent resistance is 1/n-th of the single cell’s
The larger the die, the lower its "on" resistance but at the same
time the larger its parasitic capacitances and therefore the poorer its
If everything is so exactly proportional and
predictable, is there any way for improvement?
Yes, and the idea is to
minimise (scale-down) the size of the basic cell - this way more cells can fit
in the same footprint driving the Rds(on) down while maintaining the
The continuous technological improvement and refinement of the
wafer fabrication processes (finer line lithography, better controlled implants,
etc.) is responsible for the successively improved MOSFET product generations.
But continuous striving for better processing technology is not the only way to
improve the MOSFET; conceptual design changes can result in major performance
Such a breakthrough was acheived by PHILIPS
with the development of
the TRENCH MOS process.
The gate structure, instead of being parallel to the
die surface, is now built in a trench, perpendicular to the surface, taking much
less space and making the current flow truly vertical (see fig.3).
transistors offer 50% size reduction for the same Rds(on) or a 35% size
reduction maintaining the same current handling capability.
Micro-Note helped us refresh our knowledge; we compared the MOSFET to its more
known and more used relative, the bipolar transistor. We saw the major
advantages the MOSFET has over the BJT.
We are also now aware of some
The most important conclusion is that overall circuit efficiency
is application-specific; one has to closely estimate the balance of conduction
and switching losses under all operating conditions and then decide on the
device to be used: a regular bipolar or a MOSFET.